Argument relation identification (ARI) aims to recover the directed structure of argumentative discourse, yet most approaches, including recent large language models, treat relations as local decisions and rely on implicit reasoning to enforce global coherence. This often leads to structurally inconsistent argument graphs. We reframe ARI as a global structural reasoning problem and introduce topological ordering as an explicit prediction target. Extensive experiments show consistent improvements in ARI performance, edge-level accuracy, and global structural fidelity across models and learning paradigms. In addition, the predicted topological orders provide faithful, verifiable reasoning traces that are internally consistent with model predictions and aligned with gold argument structure, offering an interpretable alternative to free-form chain-of-thought explanations.

Argument mining (AM) traditionally relies on multi-stage pipelines that train separate models for each sub-task, making end-to-end AM challenging and often unstable under multitask learning. In this work, we revisit multitask end-to-end AM in the context of large language models (LLMs) and show that multitask learning behaves fundamentally differently from pre-LLM settings. Across multiple datasets and LLMs, multitask fine-tuning of a single LLM consistently outperforms single-task models, while also exhibiting greater robustness to noisy labels and stronger cross-domain generalization. To understand these qualitatively different behaviors of LLMs, we conduct a layer-wise representation analysis, showing that cross-task features emerge in middle layers, early layers are highly domain-sensitive, and richer middle-layer representations drive performance improvements, robustness, and generalization. We further show that this mechanism extends beyond AM by observing similar patterns across other tasks. Our findings provide new insights into multitask learning with LLMs, highlight its potential for unified, efficient, and robust language understanding, and motivate layer-wise parameter efficient adaptation strategies within and beyond AM.

Argument summarization requires capturing semantic content while preserving the underlying argumentative structure. We present ArgSum-X, a dataset supporting both argumentative text summarization (ATS) and key point analysis (KPA) . To construct the dataset, we propose constructed via a multi-stage annotation framework with hybrid LLM–human annotation and quality tiers, and provide a detailed analysis of annotation quality. We propose a comprehensive evaluation framework for argument summarization covering structural, faithfulness, and semantic aspects, and conduct an empirical analysis of structure-aware modeling, revealing the unique properties and challenges of argument summarization.

Large language models (LLMs) have advanced argument mining (AM), yet most approaches treat component classification, relation identification, and explanation generation as separate tasks, limiting coherence and explainability. We propose the first explainable end-to-end AM framework grounded in Toulmin’s theory, jointly predicting argument components, argumentative relations, and structured explanations in a single generation step. This unified formulation encourages LLMs to produce internally consistent argument graphs and reasoning traces aligned with their predictions. Further, we introduce five explainability criteria tailored to AM and grounded in Toulmin’s theory: completeness, correctness, faithfulness, soundness, usefulness, and develop a unified evaluation protocol combining structural metrics, neural similarity, LLM-based judgment, human evaluation and qualitative analysis. Extensive evaluation on several LLMs shows that our approach achieves competitive performance while substantially improving explainability, generating Toulmin-based explanations that are complete, grounded and faithful, uncovering reasoning errors not captured by standard AM metrics, and being consistently preferred in usefulness evaluations.

Argument summarization seeks to generate concise representations of argumentative content while preserving argumentative meaning, yet the field remains conceptually fragmented. In this survey, we provide a unified task definition of argument summarization that encompasses both argumentative text summarization and key point analysis, and introduce a taxonomy along input structure, output representation, and the role of argumentative structure. Using this framework, we analyze the evolution of the field and systematically review existing datasets, methods, and evaluation practices, revealing misalignment that hinders comparability and progress. We further situate argument summarization within the broader context of generic text summarization, identifying practical limitations and intrinsic challenges. Finally, we outline promising future research directions.

Sentiment Analysis, a popular subtask of Natural Language Processing, employs computational methods to extract sentiment, opinions, and other subjective aspects from linguistic data. Given its crucial role in understanding human sentiment, research in sentiment analysis has witnessed significant growth in the recent years. However, the majority of approaches are aimed at the English language, and research towards Arabic sentiment analysis remains relatively unexplored. This paper presents a comprehensive and contemporary survey of Arabic Sentiment Analysis, identifies the challenges and limitations of existing literature in this field and presents avenues for future research. We present a systematic review of Arabic sentiment analysis methods, focusing specifically on research utilizing deep learning. We then situate Arabic Sentiment Analysis within the broader context, highlighting research gaps in Arabic sentiment analysis as compared to general sentiment analysis. Finally, we outline the main challenges and promising future directions for research in Arabic sentiment analysis.

As an effective way to address data scarcity problem, data augmentation has received significant interest in low-resource neural machine translation, while the latter has the potential to reduce digital divide and benefit out of domain translation. However, the existing works mainly focus on how to generate the synthetic data, while the synthetic data quality and the way we use the synthetic data also matter. In this paper, we give a systematic analysis of data augmentation for low-resource neural machine translation that encompasses all the three aspects. We show that with careful control of the synthetic data quality and the way we use the synthetic data, the performance can be greatly boosted even with the same method to generate the synthetic data.